Supply Chain LabForklift RL · Learn by rewardLab Home
Concept game · reinforcement learning

Forklift RL Playground

Teach a forklift to collect a priority pallet and deliver it to dispatch. Every move exposes the full reinforcement-learning loop: observe a state, choose an action, receive a reward and update what the agent believes.

1 · ObserveThe grid position and whether the forklift carries a pallet form the state.
2 · ActMove manually or let the agent explore and exploit its Q-values.
3 · Receive a rewardFast, safe delivery earns more; delay, collisions and hazards cost points.
4 · UpdateQ-learning shifts one state-action value toward the observed outcome.

Western Melbourne fulfilment cell

The forklift moves only when the simulated state changes.

Episode ready
Live warehouse reinforcement-learning gridA forklift learns a route between its start point, a priority pallet rack and the dispatch dock while avoiding blocked and hazardous cells.

Swipe the grid sideways to see all of it.

1Episode
0Episode return
0Successful deliveries
0%State-actions learned

Last learning step

Ready. Choose an action or ask the agent to move.

Experience stream

  • Episode 1. Find the pallet, then reach dispatch.

Current state-action values

How to read the arrows

Each warehouse cell shows the best learned action for the current load state. Red means the strongest action. Faint arrows are uncertain or have never been tried.

Current reward signal

Reward is the objective, not an instruction

StateWhere am I? Carrying?
→
ActionWhich move?
→
RewardHow useful?
→
UpdateWhat changed?

Changing the reward design changes the behaviour the same algorithm learns. That is why objective design is a management decision.

Return by episode

Learning briefing

About this game

What RL is

Reinforcement learning improves a policy through interaction. It does not need a labelled “correct action” for every state; it needs consequences.

Explore versus exploit

Exploration tests uncertain actions. Exploitation chooses the best action currently known. Epsilon controls that trade-off.

Supply-chain insight

Warehouse control, replenishment and dispatch sequencing all involve repeated decisions where today’s move changes tomorrow’s state.